PhD Projects

Over the last 40 years, a considerable number of formalisms and simulators have been developed for modeling and simulating Discrete Event Systems (DES). One formalism that has received particular attention in the research community is the Discrete Event System Specification (DEVS). This approach, today known as Classic DEVS, was developed by Zeigler from systems theory and published in 1976. Over the years, there have been further developments of the formalism with respect to specific application needs, such as fuzzy, hybrid or real-time simulations. The main evolution for general use is Parallel DEVS (PDEVS), which was introduced by Chow in 1994. Classic DEVS as well as PDEVS can be viewed primarily as extensions of a Moore machine. Therefore, modeling of certain Mealy-type components can be difficult with these formalisms. Preyser et al. analyzed this problem and presented a revised version RPDEVS in 2018. In 2020 Junglas showed that RPDEVS still has problems with complex structures of concurrent events. He proposed the use of concepts of Nonstandard Analysis to develop a more robust method of handling concurrent events. The short name of the new formalism variant should be NSA-DEVS.

Subsequently, this idea was further developed as joint research with the CEA group. Moreover, NSA-DEVS became a central component in this cooperative PhD project with the Institute for Automation at the University of Rostock.

The System Entity Structure (SES) is a high level approach for variability modeling, particularly in simulation engineering, which is under continuous development. In this context, an enhanced framework is used that supports dynamic variability evolution with the SES approach.

An SES describes a number of structure variants encoded in a tree structure with nodes and edges. On the SES a pruning operation is defined, which resolves all decision points on execution, transferring the SES in a Pruned Entity Structure (PES). The PES describes one possible structure variant. Leaf nodes can contain links to a Model Base (MB) storing basic and coupled models. With the help of a build method, an executable simulation model (SM) can be built from a PES and basic models from the MB. For automatic derivation of PES and generation, simulation, and evaluation of SMs the SES/MB approach was extended.

For this extended SES/MB (eSES/MB) framework software tools are developed. This project focuses on the development of different approaches for the generation of SMs for several simulators. In this context the Functional Mock-up Interface (FMI) as general simulator interface is connected to the eSES/MB framework.

Robots have been established in industry for decades as powerful and flexible tools. With new application areas and requirements, such as those defined in the context of the Industry 4.0 initiative, the efficient development of robot controllers is becoming increasingly important. This trend requires the development of new methods for the
manufacturer-independent and application-independent programming of robot applications. The decisive factors here are systematic procedural models, theory-compliant and modelbased development methods, and modern programming systems as development tools. The large number of manufacturer-specific development environments on the market shows that existing standards in the area of control development for robots have hardly been able to establish themselves. The methodological and software diversity makes it difficult to group different robot types and robots from different manufacturers in teams in order to implement powerful, flexible and cost-effective multi-robot systems (MRS). Continuous tool
chains as well as methods of modeling and simulation play an important role in this context.  In the literature, rapid control prototyping (RCP) is spoken of in this context. Currently, RCP-based techniques in industrial robotics are almost always manufacturer-specific.

In this thesis, an approach for end-to-end, model-based and vendor-independent control development for robot teams with industrial jointed-arm robots is developed. Based on the Simulation-Based-Control (SBC) approach and the Task-Oriented-Control (TOC) approach, development methods from the field of single robot systems (SRS) are transferred
to MRS. Possible interactions between robots are investigated and, based on this, interaction classes are defined. For the implementation of an integrated control development the Discrete-Event-System-Specification (DEVS) formalism is discussed and extensions for real-time and process integration are investigated. Derived from this, a modified formalism is developed and its suitability for integrated, model-based control development is demonstrated by means of a case study. For the specification of the developed models the DEVS diagram notation is used and extended by additional descriptive means. Based on the previously defined interaction classes, approaches for the TOC-based implementation of interactions are developed using modular and reusable tasks. Subsequently, their prototypical implementation is shown on the basis of case studies using the newly developed DEVS formalism. The complexity of the interactions increases with each case study. For a better handling of the complexity, the Extended System-Entity-Structure/Model-Base (SES/MB) approach is introduced as an additional model-based technique, integrated with the SBC and TOC approaches, and compared with the previous approach using a case study.

Modeling and Simulation (M&S) forms the basis of the modern development and planning
process in engineering. Modeling deals with the abstraction of essential aspects of technical
systems and their implementation in a model. During simulation the implemented model is
executed aiming to gain knowledge about the system behavior. A single model execution is
called a simulation run. A simple simulation experiment comprises many simulation runs
with varied input variables. Complex simulation-based experiments are characterized by the
integration of simulation with numerical methods such as optimization, screening or sensi-
tivity analysis. Highly complex simulation experiments investigate the different model
structures in addition to the parameter space.
The classical M&S approach is increasingly reaching its limits in engineering. One
reason is the growing customers demand for customized products. This increases the variety
and often the complexity of technical systems as well as of the resulting models and
investigation objectives. This means that the specification and execution of simulation
experiments includes not only a set of model variants with their parameter settings, but also
combinations with numerical methods and their configuration. Furthermore, the sequence of
variants to be investigated or numerical methods to be applied often arise from previously
determined experiment results.
Accordingly, this work contributes to the development of general methods for variant
management in M&S up to the simulation experiments’ level and their automated execution.
The proposed solutions are developed based on the ASIM procedure model, the modular-
hierarchical modeling approach and the System Entity Structure/Model Base (SES/MB)
framework. The variant management is based on different phases: variant analysis,
variant formalization, variant implementation and variant generation. Furthermore, a
modular-hierarchical structuring for simulation experiments is developed. By extending the
SES/MB framework, not only model variants but complete experiment variants can be speci-
fied, implemented and generated. Moreover, a framework for automated experimentation
based on the extended SES/MB framework is proposed. Finally, the methods are presented
using an application example from the field of production and logistics.
thesis

In practical control systems, deadzone characteristics are encountered in a wide range of mechanical and electrical components, such as valves or DC servo motors. Another process that exhibits a deadzone characteristic is the three-way catalyst. Due to its oxygen storage ability, the normalized air-fuel ratio post catalyst remains at one despite air-fuel mixture variations pre catalyst, as long as certain levels of the oxygen storage state are not exceeded. 
The aim of this project is to reduce the exhaust emissions produced by cars equipped with a three-way catalyst by improving the control of the air-fuel mixture. Low emissions are reached if post-catalyst air-fuel ratio remains at one despite large air-fuel mixture variations and a biased measurement of pre-catalyst air-fuel ratio. Typically, the aim of a controller for deadband processes is to overcome the deadzone in an optimal way, commonly by employing the deadzone inverse to cancel it. For this project, however, the aim of the controller is to keep the oxygen storage state within the deadzone. 
In order to control the oxygen storage state (and pre-catalyst air-fuel ratio which is connected to the storage state via the deadzone), a model-based approach is proposed, since the storage state cannot be measured directly by a sensor. Thus, the project includes i) the development of a control-oriented model of a three-way catalyst, which has to be both accurate and simple enough to be calculated on the vehicles electronic control unit, and ii) the design of a suitable control strategy. 

Due to increasing demands on driving comfort, consumption and emission of modern combustion engines new combustion processes came to introduction of series production in the last years. The potential of these combustion processes in general can only be utilized by introduction of additional actuating and measure variables. Thereby an increasing complexity from control viewpoint is almost unavoidable. The choice of particular actuating variables has decisive influence on efficiency, comfort and waste gas emission. Besides the main manipulated variables, as e.g. the throttle position, there are some more different actuating variables that are in many cases only effective for limited operating ranges of the engine with different effects.
The cooperative graduation project deals with the general problem of control with auxiliary actuating and control variables. Application emphasis of the procedures to be developed is located in the field of combustion engine controllers.
One aim is to find the best compromise between efficiency, driving comfort and waste gas emission. This is achieved by an optimized choice, according to several criteria, of available actuating variables for the realization of the driving torque. 
thesis

The coordination of available actuators in modern engine control units (ECUs) is a challenging task. The broad variety of signals (e.g. throttle, advance angle, exhaust gas recirculation, injection, etc.) to influence the momentum and engine speed are coupled. Therefore a multi-variable control should manage these actuators to fulfill the control aim in an efficient manner. A compromise respecting further aims like comfort issues and exhaust gas emissions must be found. An available scheme to cope with these requirements is model predictive control (MPC). The incorporated optimization ensures the optimal selection of actuator signals under their constraints. 
The cooperative graduation project is concerned with the development of a suitable quadratic program that solves the optimization within the MPC algorithm. It needs to fulfill the real-time requirements when run on high-potential micro-controllers (e.g. Infineon Tricore) that are incorporated in modern ECUs. 

Progressive robotics research opens up new application fields incessantly. Hence, demands on the development of robot controls are increasing. Easy programming and integration of different external hardware are of particular importance. The aim of every control implementation is to realise easy, safe, fast and cost-effective design and commissioning of robot applications. Therefore, it is essential to avoid re-implementations in the entire development process. 
An approach for integrated modeling, simulation and operation, named "simulation model-based rapid control prototyping", is introduced and illustrated by the example of a sensor based robot control. It is discussed how simulation models have to be structured in the early system design stage in order to extend them to model-based control programs for the operation stage stepwise. 
Objects of this research are (i) the prevention of re-implementations, (ii) the development of a task oriented programming approach, (iii) methods for an easy integration of different hardware and (iv) a concept for specification of high flexible and re-configurable controls. 
thesis

The research reported in this thesis details a new simulation based approach providing automatic reconfiguration and optimisation of both model structure and model parameters.This is achieved through a combination of simulation, optimisation and model management methods. Simulation is used to determine current model performance and an optimisation method, assistet by model management, searches for an optimal solution with repeated model parameter and model structure changes. The approach employs a meta-modeling method to define a set of model structure variants and includes a model base with pre-defined basic components. With the meta-modeling method the model management can determine specific model structures and create executeable models.

Discrete Event System Simulation (DEVS) in our days is not widely accepted by engineers. This is mainly caused by the fact that engineers usually are rather familar with Scientific and technical Computing Environments (SCEs) than with the use of high level programming language simulation libraries. Furthermore, most of those comercial off-the-shelf tools are just suitable for specific application areas.
One more aspect is that SCEs offer good opportunies to integrate DEVS modeling and simulation with other advanced techniques such as e.g. optimisation, ode solvers, data aquisition and analysis and integration of hardware.
In this project main focus is put on modeling and simulation of hybrid system dynamic by utilisation of the features for solving ordinary differential equations within the most popular SCE Matlab. The approach includes modeling of structure variable systems, as well.

Scientific and technical Computing Environments (SCEs) like Matlab are powerful tools for todays engineers, especially for desing problems. But the increasing complexity of calculations often lead to bottlenecks in interactive design processes. Parallel processing offers one possibility to bypass those bottlenecks. Within this research project, approaches for combining SCEs and parallel processing are investigated and classified. Several tools are examined with respect to their programming model, abstraction level and communication performance. Furthermore, several example programs are parallelized for runtime measurements and implementation effort investigations.

Kinematic robot calibration is the key requirement for the successful application of offline programming to industrial robotics. To compensate for inaccurate robot
tool positioning, offline generated poses need to be corrected using a calibrated kinematic model, leading the robot to the desired poses. Conventional robot calibration techniques are heavily reliant upon numerical optimisation methods for model parameter estimation. However, the non-linearities of the kinematic equations, inappropriate model parameterisations with possible parameter discontinuities or redundancies, typically result in badly conditioned parameter identification. Research in kinematic robot calibration has therefore mainly focused on finding robot models and appropriate accommodated numerical methods to increase the accuracy of these models. 

This thesis presents an alternative approach to conventional kinematic robot calibration and develops a new inverse static kinematic calibration method based on
the recent genetic programming paradigm. In this method the process of robot calibration is fully automated by applying symbolic model regression to model
synthesis (structure and parameters) without involving iterative numerical methods for parameter identification, thus avoiding their drawbacks such as local convergence, numerical instability and parameter discontinuities. The approach developed in this work is focused on the evolutionary design and implementation of
computer programs that model all error effects in particular non-geometric effects such as gear transmission errors, which considerably affect the overall positional accuracy of a robot. Genetic programming is employed to account for these effects and to induce joint correction models used to compensate for positional errors. The potential of this portable method is demonstrated in calibration experiments carried out on an industrial robot.